I’m going to respond to what Minh said here about the US (and to some degree Canadian) context. First, private crossings indeed are a particular concern for us, as the normal private crossing signage is intended to substitute for the crossbuck in most cases, and may vary state by state. Furthermore, not all states require trains to sound their horns for private crossings under the presumption that private crossings are only utilized by authorized users who are aware of the crossing’s presence.
(Note that there is a minority of private crossings with crossbucks in the US, and even some that have active warning systems – note also that the private crossing sign itself isn’t mandatory at private crossings, as seen in the last example, which provides access to a former military site that’s now a nature preserve.)
The more pressing matter, though, is that street running rail in this country takes two forms, and is accompanied by a third form of embedded trackage that is found in the off-street environment. The first form is what I call “mixed-traffic street running” where the train tracks are in a travel lane, as seen on Virtual Railfan’s La Grange, KY camera. This is tagged using the embedded key on the train tracks and the embedded_rails key on the road, with embedded_rails:lanes used for detailing what lanes the train tracks can be found in.
The second form of street running is what I call “median running”, where cars are not permitted to use the lanes occupied by rails. (This can also happen in a “mounded” situation where the tracks, even though they are within the paved limits of the street, are not flush with the pavement, but built atop the pavement, but that isn’t a concern here.) The classical median-running example is Jack London Square in Oakland, CA, which is also some of the busiest street-running track in the US. We map this using embedded on the railway, but without using the embedded_rails tags on the highway ways, as they are split around the tracks, just as if one is mapping a divided highway.
Finally, we have the situation where embedded trackage is found outside street rights of way. This happens in intermodal and automotive terminals, as well as at ports and some other industrial facilities where road vehicles (trucks) and heavy equipment must have close access to railcars. Once again, we should map this with embedded on the tracks in question; however, it is unclear if embedded_rails is appropriate to use on paved “apron” areas given that tracks in an apron can and do take non-straightforward paths. It’s also notable that most of these situations do not have saltires or other crossing signage present; in fact, oftentimes, and unlike the other two cases above, there may only be a single “blanket” record in the FRA (federal) database for all conflict points within the facility, or even no crossings at all within the apron, legally speaking. While these areas are not open to the public, some OSM users (such as truckers) will encounter these crossings, and the risks involved with industrial crossings can be quite high despite the slow speeds, as switching (shunting) moves often require railroad personnel to be in positions where they do not have the protection from impact granted by a locomotive’s cab.
As a result of what I said above, this, however, isn’t correct for the way we map street running mainline situations, as there are two sorts of crossing nodes in this case. The first sort are the nodes mapped on the cross streets, which may very well have saltires or even active warnings, such as lights and crossing gates. Furthermore, there is nothing stopping street-running railway track from being part of a quiet zone, where trains need not sound their horns unless the train crew spies an idiot playing chicken with them. (Indeed, the La Grange street running is within a quiet zone.)
However, there may also be crossing nodes at the entrances to street running sections, and the verdict on those is…more variable. It is plausible that a short length of street running industrial track might not have warning devices at the entrances, relying on highway traffic signal preemption or even train crew members getting off to stop the road traffic. However, it’s also possible for these entrances to take place at full-fledged crossings, often within the context of an intersection.
And that’s before we get into LRT, where tram-like segments where LRVs treat themselves equivalently to road vehicles are mixed with railroad-like segments, some of which occupy former railroad rights of way (Dallas’ DART system is known to push 70mph on the latter). This means that railway=light_rail could follow either tram rules or train rules.
So, whatever we do in the future:
- shouldn’t apply to heavy rail, almost ever. Neither the saltire-heuristic (due to private crossings) nor the embedded-heuristic (due to mainline street running) works in this case – it may simply be better to treat
tram_crossing on mainline rails as flatly deprecated/erroneous tagging, even.
- Needs to distinguish cases where light rail/trams run in mixed traffic (as road vehicles, running “on sight”) from when they run as rail vehicles (in separate ROW, usually on signal indication, and with a presumption that grade crossing treatments are applied even when the line engages in median or street running under that circumstance).
- Needs to understand that extreme corner cases (such as SF’s Cargo Way Interlocking, where rail signals govern a crossing between median-running LRT and a heavy rail industrial spur) do exist.
- Probably shouldn’t apply to miniature railways either (these are in parks or other contexts where a street ROW would indeed be interacting with them as a crossing, not in the form of street running).
- And needs to be intuitive enough to work across a wide variety of countries and transit traditions.
@Minh_Nguyen can you come up with a case on a railway=light_rail here in the US where the LRVs run on signal indication (vs. “on sight”) yet a saltire or other crossing protection is not present at crossings of the LRT line?
